ObjView-Bench: Rethinking Difficulty and Deployment for Object-Centric View Planning

This paper introduces ObjView-Bench, an evaluation framework designed to address the limitations of existing object-centric view planning assessments. The authors disentangle three key quantities: omnidirectional self-occlusion as an object-side attribute, observation saturation difficulty, and protocol-dependent planning difficulty defined via a set-cover formulation. By separating these factors, the framework supports controlled dataset construction and analysis of slow-saturation objects. Additionally, the study designs deployment-oriented evaluation protocols to reveal how budget regimes and reachable-view constraints affect method behavior. Results across classical, learned, and hybrid planners demonstrate that accounting for difficulty, budget, and reachability significantly alters method rankings and failure modes.